Hi Minh,
Why not free() inside mdtm_sendto() and mdtm_mcast_sendto()?
It will help reduce much code change.
Best Regards,
ThuanTr
-Original Message-
From: Minh Chau
Sent: Tuesday, November 26, 2019 7:02 PM
To: thuan.t...@dektech.com.au; vu.m.ngu...@dektech.com.au;
But after all retries are still failed, we might need to terminate the
portid, which leads to a MDS DOWN event, but let's look at it later.
On 27/11/19 3:23 pm, Minh Hon Chau wrote:
Hi Thuan,
I'm thinking to retry 3 times with 100 ms in between, but you can
decide it. Also, we need to ensure
Hi Thuan,
I'm thinking to retry 3 times with 100 ms in between, but you can decide
it. Also, we need to ensure not to make the mds main receiving thread
being blocked with the retry (on the flow of processing data). The retry
in this patch is ok since it retries on the mds flow control
Hi Minh,
I think it's good if retry some times for normal Send().
Do you have any idea how many retries? Interval b/w tries?
Best Regards,
ThuanTr
-Original Message-
From: Minh Hon Chau
Sent: Wednesday, November 27, 2019 10:30 AM
To: thuan.tran ; thang . d . nguyen
; 'Nguyen Minh Vu'
Hi Thuan,
The TipcPortId:Send is also called at a few other places, do you think
it is good if we make a wrapper of TipcPortId::Send with a few retries
on failures, says TipcPortId::TryToSend(), and call TryToSend() instead
of Send()?
Thanks
Minh
On 27/11/19 1:26 pm, thuan.tran wrote:
Summary: mds: Fix mds flow control keep all messages in queue [#3123]
Review request for Ticket(s): 3123
Peer Reviewer(s): Minh, Vu, Thang, Gary
Pull request to: *** LIST THE PERSON WITH PUSH ACCESS HERE ***
Affected branch(es): develop
Development branch: ticket-3123
Base revision:
When overflow happens, mds with flow control enabled may keep
all messages in queue if it fails to send a message when receiving
Nack or ChunkAck since no more trigger come after that.
MDS flow control should retry to send message in this scenario.
---
src/mds/mds_tipc_fctrl_portid.cc | 16
Summary: mds: not waste 1.5s in waiting dead Adest to send RSP [#3102]
Review request for Ticket(s): 3102
Peer Reviewer(s): Minh, Vu, Thang, Gary
Pull request to: *** LIST THE PERSON WITH PUSH ACCESS HERE ***
Affected branch(es): develop
Development branch: ticket-3102
Base revision:
- When sending response message to Adest which is not exist (crash/terminate),
current MDS try to wait for 1.5 seconds before conclude no route to send RSP.
- Here are scenarios may waste 1.5s waiting:
SVCs DOWN (dead adest or vdest role change) -> get SNDRSP -> send RSP (wait
1.5s)
get
Summary: mds: Avoid message reallocation V2 [#3089]
Review request for Ticket(s): 3089
Peer Reviewer(s): Thuan, Vu, Gary
Pull request to: *** LIST THE PERSON WITH PUSH ACCESS HERE ***
Affected branch(es): develop
Development branch: ticket-3089
Base revision:
The patch avoids message reallocation if the message is in
retransmission queue
---
src/mds/mds_dt_tipc.c| 42 +++-
src/mds/mds_tipc_fctrl_intf.cc | 6 --
src/mds/mds_tipc_fctrl_intf.h| 4 ++--
src/mds/mds_tipc_fctrl_msg.cc| 2 +-
Correct indent and reduce code lines (<80 chars) for
mds_mdtm_send_tipc() and mdtm_frag_and_send()
---
src/mds/mds_dt_tipc.c | 484 ++
1 file changed, 254 insertions(+), 230 deletions(-)
diff --git a/src/mds/mds_dt_tipc.c b/src/mds/mds_dt_tipc.c
12 matches
Mail list logo